Update app.py

app.py

@@ -1,12 +1,16 @@
 import gradio as gr
-from datasets import load_dataset
+from datasets import load_dataset, Features, Audio
+
+# --- Revised Code ---
+# It's better to load a small dataset directly without streaming.
+# This is more stable and loads the data into memory for quick access.
+try:
+    dataset = load_dataset("Cnam-LMSSC/vibravox-test", "speech_clean", split="train")
+except Exception as e:
+    dataset = None
+    app_error = e
+# --------------------
 
-# Load the dataset from the Hugging Face Hub
-# Using streaming=True makes it load faster as it doesn't download everything at once.
-dataset = load_dataset("Cnam-LMSSC/vibravox-test", "speech_clean", split="train", streaming=True)
-# Convert to an iterable to easily access rows by index
-iterable_dataset = iter(dataset)
-rows = list(iterable_dataset)
 
 # Define the audio columns we want to display
 AUDIO_COLUMNS = [
@@ -23,16 +27,13 @@ def get_audio_row(index):
     This function retrieves a specific row from the dataset by its index.
     It returns the audio data for each of the specified audio columns.
     """
-    # Gradio takes index as a float from the slider, so we convert it to an integer
     row_index = int(index)
-    # Get the specific row from our list of rows
-    sample = rows[row_index]
+    # Get the specific row from the dataset
+    sample = dataset[row_index]
     
-    # Extract the sentence and all audio file data
     sentence = sample["sentence"]
     
     # Return the sentence and one audio object for each column
-    # The order of returned values must match the order of the gr.Audio outputs below
     return [
         sentence,
         sample["audio.headset_microphone"]["path"],
@@ -46,37 +47,46 @@ def get_audio_row(index):
 # Build the Gradio Interface
 with gr.Blocks() as demo:
     gr.Markdown("# Vibravox Multi-Audio Viewer")
-    gr.Markdown("Select a row from the dataset to listen to all corresponding audio sensor recordings.")
-    
-    # Input slider to select the row number
-    slider = gr.Slider(minimum=0, maximum=len(rows) - 1, step=1, value=0, label="Select Data Row")
     
-    # Output for the sentence
-    sentence_output = gr.Textbox(label="Sentence", interactive=False)
-    
-    # This is the key part: A horizontal row for all audio players
-    with gr.Row():
-        audio_headset = gr.Audio(label="Headset Mic", type="filepath")
-        audio_throat = gr.Audio(label="Throat Mic", type="filepath")
-        audio_soft_ear = gr.Audio(label="Soft In-Ear Mic", type="filepath")
-    with gr.Row():
-        audio_rigid_ear = gr.Audio(label="Rigid In-Ear Mic", type="filepath")
-        audio_forehead = gr.Audio(label="Forehead Accel.", type="filepath")
-        audio_temple = gr.Audio(label="Temple Pickup", type="filepath")
+    # --- Revised Code ---
+    # Handle the case where the dataset fails to load
+    if dataset is None:
+        gr.Markdown(f"## 💥 Application Error")
+        gr.Markdown(f"Could not load the dataset. This is likely due to a missing dependency. Please ensure `requirements.txt` is correct. Error: `{app_error}`")
+    else:
+        gr.Markdown("Select a row from the dataset to listen to all corresponding audio sensor recordings.")
+        
+        slider = gr.Slider(minimum=0, maximum=len(dataset) - 1, step=1, value=0, label="Select Data Row")
+        
+        sentence_output = gr.Textbox(label="Sentence", interactive=False)
+        
+        with gr.Row():
+            audio_headset = gr.Audio(label="Headset Mic", type="filepath")
+            audio_throat = gr.Audio(label="Throat Mic", type="filepath")
+            audio_soft_ear = gr.Audio(label="Soft In-Ear Mic", type="filepath")
+        with gr.Row():
+            audio_rigid_ear = gr.Audio(label="Rigid In-Ear Mic", type="filepath")
+            audio_forehead = gr.Audio(label="Forehead Accel.", type="filepath")
+            audio_temple = gr.Audio(label="Temple Pickup", type="filepath")
+
+        # This makes the UI load the first row automatically on startup
+        demo.load(
+            fn=get_audio_row,
+            inputs=gr.State(0), # Start with index 0
+            outputs=[
+                sentence_output, audio_headset, audio_throat, audio_soft_ear,
+                audio_rigid_ear, audio_forehead, audio_temple
+            ]
+        )
 
-    # Connect the slider input to the function and the outputs
-    slider.change(
-        fn=get_audio_row, 
-        inputs=slider, 
-        outputs=[
-            sentence_output,
-            audio_headset,
-            audio_throat,
-            audio_soft_ear,
-            audio_rigid_ear,
-            audio_forehead,
-            audio_temple
-        ]
-    )
+        slider.change(
+            fn=get_audio_row, 
+            inputs=slider, 
+            outputs=[
+                sentence_output, audio_headset, audio_throat, audio_soft_ear,
+                audio_rigid_ear, audio_forehead, audio_temple
+            ]
+        )
+    # --------------------
 
 demo.launch()